摘要
针对高强度噪声图像,提出了一种新的基于信息测度概念和Dempster-Shafer(DS)证据理论的边缘检测算法.利用邻域一致性、方向性和结构性3种信息测度定量描述边缘特征;引入检测不确定性,根据各信息测度响应分布设计基本可信度分配函数,并利用DS合成规则加以融合;融合后根据组合决策规则将像素分类成边缘与非边缘.实验通过检测结果以及Pratt品质因数的分析比较,表明该算法能够有效地区分边缘点和噪声点.在低噪声情况下,检测性能与传统检测方法相近;而对于高强度噪声图像,该方法具有较强的噪声免疫力.
A novel edge detection method based on information measure concept and Dempster-Shafer（DS） evidence theory was proposed for high noise image. Three information measures including neighborhood homogeneity information measure, direction information measure and structure information measure were used to characterize edge. Edge detection uncertainty was introduced to design mass functions based on the distribution of each information measure response and combine them by DS combination rule. Then pixels were classified into edge or off-edge according to combining decision rule. Detection results and Pratt figure-of-merit comparision show that the method can effectively distinguish between edge dots and noise dots and it has the same performance to classical edge detectors for low noise image, while offers superior denoising performance for high noise image.
出处
《浙江大学学报（工学版）》
EI
CAS
CSCD
北大核心
2008年第10期1671-1675,共5页
Journal of Zhejiang University：Engineering Science
基金
浙江省重点科研资助项目(2006C21037)
关键词
图像边缘检测
DS证据理论
信息测度
数据融合
image edge detection
Dempster-Shafer evidence theory
information measure
data fusion
